# Can the Radeon RX 9070 XT run Darwin-27B-RSI?

> Just about. At Q3_K_M it needs about 15 GB of the 16 GB available, so keep the context short, at roughly 27 tokens/s.

Radeon RX 9070 XT: 16 GB (16 GB usable), 640 GB/s. Estimates at an 8K context.

## Every quantisation

| Quant | Quality | ≈ GB | Fit | Est. tok/s |
|---|---|---|---|---|
| Q2_K | Heavy loss | 12.85 | fits | 40 |
| Q3_K_M (pick) | Noticeable loss | 15.37 | tight | 27 |
| Q4_K_M | Small loss | 18.94 | offload | 8.6 |
| Q5_K_M | Near-lossless | 22.72 | offload | 4.8 |
| Q6_K | Near-lossless | 25.6 | offload | 3.4 |
| Q8_0 | Lossless | 30.54 | offload | 2.2 |
| F16 | Lossless | 55.12 | no | — |

## Run it

Q3_K_M straight from Hugging Face (ollama):

```bash
ollama run hf.co/bartowski/FINAL-Bench_Darwin-27B-RSI-GGUF:Q3_K_M
```

Docker with ROCm (AMD) (ollama):

```bash
docker run -d --name ollama --device /dev/kfd --device /dev/dri \
  -v ollama:/root/.ollama -p 11434:11434 ollama/ollama:rocm
docker exec -it ollama ollama run hf.co/bartowski/FINAL-Bench_Darwin-27B-RSI-GGUF:Q3_K_M
```

OpenAI-compatible server with a web UI (llama.cpp):

```bash
llama-server -hf bartowski/FINAL-Bench_Darwin-27B-RSI-GGUF:Q3_K_M
```

Docker with ROCm (AMD) (llama.cpp):

```bash
docker run -d --name llama --device /dev/kfd --device /dev/dri -p 8080:8080 \
  -v ~/.cache/llama.cpp:/root/.cache/llama.cpp \
  ghcr.io/ggml-org/llama.cpp:server-rocm \
  -hf bartowski/FINAL-Bench_Darwin-27B-RSI-GGUF:Q3_K_M --host 0.0.0.0 --port 8080 -ngl 99
```

Download with the lms CLI (lmstudio):

```bash
lms get bartowski/FINAL-Bench_Darwin-27B-RSI-GGUF@Q3_K_M
```

Serve an OpenAI-compatible API (vllm):

```bash
vllm serve FINAL-Bench/Darwin-27B-RSI
```

Docker Compose (AMD ROCm) (vllm):

```yaml
services:
  vllm:
    image: rocm/vllm:latest
    command: vllm serve FINAL-Bench/Darwin-27B-RSI --host 0.0.0.0 --port 8000
    ports:
      - "8000:8000"
    ipc: host
    devices:
      - /dev/kfd
      - /dev/dri
    group_add:
      - video
    security_opt:
      - seccomp=unconfined
    volumes:
      - ~/.cache/huggingface:/root/.cache/huggingface
```

Serve an OpenAI-compatible API (sglang):

```bash
sglang serve FINAL-Bench/Darwin-27B-RSI --host 0.0.0.0 --port 30000
```

Chat with the pipeline API (transformers):

```python
from transformers import pipeline

pipe = pipeline("text-generation", model="FINAL-Bench/Darwin-27B-RSI", dtype="auto", device_map="auto")
messages = [{"role": "user", "content": "Explain self-hosting in one paragraph."}]
out = pipe(messages, max_new_tokens=512)
print(out[0]["generated_text"][-1]["content"])
```


## Chat models that fit comfortably

- [openai/gpt-oss-20b](https://selfhost.directory/models/openai/gpt-oss-20b/on/rx-9070-xt): Q8_0 ≈12.72 GB, est. 79 tok/s
- [Qwen/Qwen3-14B](https://selfhost.directory/models/qwen/qwen3-14b/on/rx-9070-xt): Q6_K ≈13.6 GB, est. 29 tok/s
- [microsoft/phi-4](https://selfhost.directory/models/microsoft/phi-4/on/rx-9070-xt): Q6_K ≈13.82 GB, est. 29 tok/s
- [Qwen/Qwen2.5-14B-Instruct](https://selfhost.directory/models/qwen/qwen2.5-14b-instruct/on/rx-9070-xt): Q6_K ≈13.85 GB, est. 29 tok/s
- [Qwen/Qwen3-8B](https://selfhost.directory/models/qwen/qwen3-8b/on/rx-9070-xt): Q8_0 ≈10.15 GB, est. 40 tok/s
- [meta-llama/Llama-3.1-8B-Instruct](https://selfhost.directory/models/meta-llama/llama-3.1-8b-instruct/on/rx-9070-xt): Q8_0 ≈9.84 GB, est. 41 tok/s

## Other GPUs that run it

- [Tesla T4](https://selfhost.directory/models/final-bench/darwin-27b-rsi/on/tesla-t4): Q3_K_M, est. 13 tok/s
- [Arc A770 16GB](https://selfhost.directory/models/final-bench/darwin-27b-rsi/on/arc-a770): Q3_K_M, est. 23 tok/s
- [Radeon RX 7800 XT](https://selfhost.directory/models/final-bench/darwin-27b-rsi/on/rx-7800-xt): Q3_K_M, est. 26 tok/s
- [Tesla P40](https://selfhost.directory/models/final-bench/darwin-27b-rsi/on/tesla-p40): Q5_K_M, est. 10 tok/s
- [Radeon RX 7900 XTX](https://selfhost.directory/models/final-bench/darwin-27b-rsi/on/rx-7900-xtx): Q5_K_M, est. 28 tok/s
- [Instinct MI50 32GB](https://selfhost.directory/models/final-bench/darwin-27b-rsi/on/instinct-mi50): Q6_K, est. 25 tok/s
- [GeForce RTX 5090](https://selfhost.directory/models/final-bench/darwin-27b-rsi/on/rtx-5090): Q6_K, est. 44 tok/s
- [Apple M4 Pro (48 GB)](https://selfhost.directory/models/final-bench/darwin-27b-rsi/on/apple-m4-pro-48gb): Q8_0, est. 5.3 tok/s

---
Model: https://selfhost.directory/models/final-bench/darwin-27b-rsi. GPU: https://selfhost.directory/models/gpus/rx-9070-xt. Source: https://selfhost.directory/models/final-bench/darwin-27b-rsi/on/rx-9070-xt